Condensation monitoring method of digital substation

Abstract

The invention discloses a condensation monitoring method for a digital substation. The method comprises the following steps: acquiring data information of a target digital substation; a high-risk area where condensation occurs is determined, and monitoring equipment is arranged to obtain monitoring data; selecting a monitoring frequency band, acquiring and analyzing data of the monitoring frequency band, and constructing a feature set related to the condensation degree; acquiring monitoring data of the target digital substation in real time and selecting an electromagnetic wave monitoring frequency band; feature extraction and analysis are carried out to construct a condensation monitoring model, and optimization is carried out; generating a condensation risk prediction result of the target digital substation based on the optimized condensation monitoring model; and giving an optimal anti-condensation measure for the target digital substation according to a prediction result, and completing condensation monitoring of the target digital substation. According to the method, the environmental parameters of the digital substation are acquired and processed, and the condensation prediction model is established, so that the condensation monitoring of the digital substation is realized, the reliability is higher, and the accuracy is better.

Description

Technical Field

[0001] The invention belongs to the field of electrical automation, and in particular relates to a condensation monitoring method for a digital substation. Background Art

[0002] With the development of economy and technology and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and life, bringing endless convenience to people's production and life. Therefore, ensuring the stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] Substations are an important part of the power system, so the working status of substations will directly affect the operating status of the power system. The internal environment of substations is complex and changeable. Changes in factors such as temperature and humidity can easily lead to condensation on the surface of equipment, affecting the insulation performance and operational reliability of the equipment; therefore, condensation monitoring in substations is of great significance.

[0004] At present, the traditional condensation monitoring solution adopted by substations is still a simple sensor + threshold alarm solution. Although this type of solution is simple, the reliability of monitoring data is poor, and the accuracy of condensation monitoring results is also low. This undoubtedly makes the condensation monitoring effect of substations poor, thus bringing great hidden dangers to the safe and stable operation of substations. Summary of the invention

[0005] The object of the present invention is to provide a condensation monitoring method for a digital substation with high reliability and good accuracy.

[0006] The condensation monitoring method for a digital substation provided by the present invention comprises the following steps:

[0007] S1. Obtain data information of the target digital substation;

[0008] S2. According to the data information obtained in step S1, based on the numerical simulation scheme, simulate the airflow movement characteristics in the target digital substation, determine the high-risk areas for condensation, and arrange corresponding monitoring equipment to obtain monitoring data;

[0009] S3. According to the difference in the response characteristics of electromagnetic waves in different frequency bands to condensation, a monitoring frequency band is selected, and data of the monitoring frequency band is obtained and analyzed, so as to construct a feature set related to the degree of condensation;

[0010] S4. Obtain the monitoring data of the target digital substation in real time, and select the electromagnetic wave monitoring frequency band based on the current environment;

[0011] S5. According to the electromagnetic wave monitoring frequency band obtained in step S4, feature extraction and analysis are performed to construct a condensation monitoring model based on electromagnetic wave transmission characteristics;

[0012] S6. Optimizing the condensation monitoring model constructed in step S5 based on the fused environmental parameters and electromagnetic wave parameters;

[0013] S7. Based on the optimized condensation monitoring model obtained in step S6, a condensation risk prediction result of the target digital substation is generated;

[0014] S8. According to the prediction result obtained in step S7, the optimal anti-condensation measures are given to the target digital substation, and the condensation monitoring of the target digital substation is completed.

[0015] The step S2 specifically includes the following steps:

[0016] A three-dimensional model of the target digital substation is established, and the airflow movement in the target digital substation is numerically simulated using a computational fluid dynamics solution to obtain the characteristic parameters of the airflow movement in different areas of the target digital substation;

[0017] The obtained air flow motion characteristic parameters are correlated with temperature and humidity to determine the high-risk areas for condensation in the target digital substation;

[0018] In the identified high-risk areas, corresponding monitoring equipment is deployed to obtain monitoring data.

[0019] The step S3 specifically includes the following steps:

[0020] Based on the electromagnetic wave transmission theory and the analysis of the physical mechanism of condensation, several frequency bands that are sensitive to condensation are selected as candidate frequency bands;

[0021] Build a condensation simulation experimental platform, conduct experiments under different condensation conditions, and obtain electromagnetic wave transmission characteristic test data of candidate frequency bands;

[0022] Fitting and regression analysis were performed on the obtained electromagnetic wave transmission characteristic test data to obtain the correlation scale model between the condensation degree and the electromagnetic wave frequency band and transmission characteristic parameters;

[0023] According to the correlation scale model, the monitoring frequency band is determined to obtain a monitoring frequency band set;

[0024] According to the obtained monitoring frequency band, electromagnetic wave transmission data is collected to obtain an electromagnetic wave transmission data set reflecting the condensation state;

[0025] Perform video domain analysis on the electromagnetic wave transmission data reflecting the condensation state, and extract characteristic parameters that can reflect the degree of condensation;

[0026] The feature selection algorithm is used to screen and optimize the obtained feature parameters, obtain the optimal feature subset related to the condensation degree, and construct the condensation degree prediction feature set;

[0027] The obtained condensation degree prediction feature set is used to train the support vector regression algorithm to obtain a condensation degree prediction model;

[0028] The obtained condensation degree prediction model is used to predict the condensation degree of the target digital substation.

[0029] The step S4 specifically comprises the following steps:

[0030] Real-time acquisition of environmental monitoring data of the target digital substation; the environmental monitoring data includes temperature data, humidity data, pressure data and light intensity data of the UB Aokuo monitoring point;

[0031] Clustering algorithm is used to perform clustering analysis on the acquired environmental monitoring data, and the target digital substation is divided into several environmental areas according to the clustering results;

[0032] Install condensation sensors in various environmental areas to monitor condensation data on the equipment surface in real time;

[0033] Select a monitoring frequency band and use the selected monitoring frequency band to transmit condensation data data.

[0034] The step S5 specifically includes the following steps:

[0035] Based on the maximum relevance minimum redundancy algorithm and the Relief-F algorithm, feature selection is performed on the electromagnetic wave attenuation feature data, and the attenuation feature with the greatest correlation with the condensation degree is screened out as the candidate feature set.

[0036] Feature extraction and fusion are performed on the candidate feature set. The principal component analysis method is used to extract the main component features of the attenuation feature. The independent component analysis method is used to extract the independent component features of the attenuation feature. The autoencoder is used to extract the deep nonlinear features of the attenuation feature to obtain the attenuation feature parameters.

[0037] Acquire condensation image data and video data on the device surface;

[0038] Extract image feature parameters of the acquired image data and video data, score the condensation degree according to the set requirements, and obtain a condensation observation sample data set;

[0039] The condensation observation sample data set is used to train the gradient boosting regression tree model to obtain the condensation monitoring model.

[0040] Obtain the weather forecast data, internal environment monitoring data and electromagnetic wave attenuation data of the target digital substation, and use the obtained data to train the long short-term memory neural network to obtain the condensation trend prediction model;

[0041] The condensation trend prediction model is used to predict the condensation trend of the target substation.

[0042] The step S6 specifically includes the following steps:

[0043] The acquired environmental parameters and electromagnetic wave parameters are preprocessed through data alignment, time series interpolation and outlier removal algorithms to obtain fused feature vectors;

[0044] The condensation monitoring model constructed in step S5 is optimized by using the fused feature vector.

[0045] The step S7 specifically includes the following steps:

[0046] The optimized condensation monitoring model is used to predict the condensation risk of the target digital substation;

[0047] According to the predicted condensation risk probability, clustering algorithm is used to perform clustering and grouping, and different risk levels are marked;

[0048] The monitoring grid of the target digital substation is set as an undirected weighted graph. Each monitoring point corresponds to a node of the graph. There is a weighted edge between adjacent monitoring points, and the weight of the edge is set. The Node2Vec algorithm is used to optimize the node collinearity probability to obtain the low-dimensional embedding vector of the node. The IDW algorithm is used to interpolate and predict the risk level of unknown nodes. Based on the interpolation prediction results, a risk distribution map is generated with the substation geographic information map as the base map and the monitoring points as control points. Areas with different risk levels are marked.

[0049] The step S8 specifically includes the following steps:

[0050] According to the prediction result obtained in step S7, the area where the condensation risk is greater than the set threshold is obtained, and the specific location and coverage are determined in combination with the layout data of the target substation to obtain a location list of high-risk areas for condensation;

[0051] Build a knowledge base of anti-condensation measures;

[0052] In the anti-condensation measures knowledge base, the environmental characteristics of the high-risk area for condensation are compared with the environmental characteristics data in the anti-condensation measures knowledge base for similarity, and the anti-condensation measures with similarity higher than the set threshold are selected as candidate measures;

[0053] Among the candidate measures, the cost data of each candidate measure is calculated according to the installation location and installation method of the candidate measure, and the changes in regional environmental parameters after the implementation of each candidate measure are calculated through the three-dimensional model of the target digital substation, and the anti-condensation effect of the candidate measures is evaluated. According to the cost data and the anti-condensation effect, the final anti-condensation measures are determined to complete the condensation monitoring of the target digital substation.

[0054] The condensation monitoring method for a digital substation further includes the following steps:

[0055] S9. Establish a condensation monitoring and protection platform to achieve condensation monitoring, data management and optimization of protection measures for the target digital substation.

[0056] The step S9 specifically includes the following steps:

[0057] Perform data preprocessing on the acquired data to generate a condensation monitoring data set;

[0058] Based on support vector machine and convolutional neural network, an initial model for condensation risk assessment is constructed, and the generated condensation monitoring data set is used for training to obtain a condensation risk assessment model.

[0059] The obtained condensation risk assessment model is used to obtain the condensation risk assessment result, and combined with the three-dimensional model of the target digital substation, a condensation risk distribution map is generated;

[0060] According to the generated condensation risk distribution map, the anti-condensation measures are dynamically optimized based on a multi-objective planning scheme.

[0061] The condensation monitoring method for a digital substation provided by the present invention not only realizes the condensation monitoring of the digital substation but also has higher reliability and better accuracy by acquiring and processing the environmental parameters of the digital substation and establishing a condensation prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 The figure is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0063] like Figure 1 The method flow chart of the method of the present invention is shown as follows: the condensation monitoring method of the digital substation disclosed in the present invention comprises the following steps:

[0064] S1. Obtain data information of the target digital substation;

[0065] S2. According to the data information obtained in step S1, based on the numerical simulation scheme, simulate the airflow movement characteristics in the target digital substation, determine the high-risk area for condensation, and arrange corresponding monitoring equipment to obtain monitoring data; specifically, the following steps are included:

[0066] A three-dimensional model of the target digital substation is established, and the airflow movement in the target digital substation is numerically simulated using a computational fluid dynamics solution to obtain the characteristic parameters of the airflow movement in different areas of the target digital substation;

[0067] The obtained air flow motion characteristic parameters are correlated with temperature and humidity to determine the high-risk areas for condensation in the target digital substation;

[0068] In the identified high-risk areas, deploy corresponding monitoring equipment to obtain monitoring data;

[0069] In the specific implementation, the computational fluid dynamics method is used to numerically simulate the air flow movement inside the substation, the finite volume method is used to discretize the control equations, and the SIMPLE algorithm is used to solve the velocity field and pressure field. The calculation domain is divided into 500,000 hexahedral grids, the boundary conditions are set as the inlet velocity of 5m / s, the outlet pressure is standard atmospheric pressure, and the wall is an adiabatic no-slip boundary. The simulation results show that the air flow velocity distribution inside the substation is uneven, the air flow velocity is low near the equipment, and the air flow velocity is high in the channel, with a maximum speed of 2m / s. The air flow velocity and direction are correlated with temperature and humidity. The least squares method is used to fit that the air flow velocity is negatively correlated with temperature and positively correlated with humidity, with correlation coefficients of -78 and 69, respectively. According to the results of the correlation analysis, the area near the equipment is determined to be a high-risk area for condensation. 10 temperature and humidity sensors and 5 electromagnetic wave monitoring points are arranged in this area with a sampling frequency of 1 minute to collect environmental parameters and electromagnetic wave intensity data in real time. Through statistical analysis of monitoring data, it is found that the average temperature in the high-risk area of ​​condensation is 25°C, the average humidity is 85%, and the average electromagnetic wave intensity is 50V / m, providing data support for condensation warning;

[0070] S3. According to the difference in the response characteristics of electromagnetic waves in different frequency bands to condensation, a monitoring frequency band is selected, and data of the monitoring frequency band is obtained and analyzed, so as to construct a feature set related to the degree of condensation; specifically, the steps include:

[0071] Based on the electromagnetic wave transmission theory and the analysis of the physical mechanism of condensation, several frequency bands that are sensitive to condensation are selected as candidate frequency bands;

[0072] Build a condensation simulation experimental platform, conduct experiments under different condensation conditions, and obtain electromagnetic wave transmission characteristic test data of candidate frequency bands;

[0073] Fitting and regression analysis were performed on the obtained electromagnetic wave transmission characteristic test data to obtain the correlation scale model between the condensation degree and the electromagnetic wave frequency band and transmission characteristic parameters;

[0074] According to the correlation scale model, the monitoring frequency band is determined to obtain a monitoring frequency band set;

[0075] According to the obtained monitoring frequency band, electromagnetic wave transmission data is collected to obtain an electromagnetic wave transmission data set reflecting the condensation state;

[0076] Perform video domain analysis on the electromagnetic wave transmission data reflecting the condensation state, and extract characteristic parameters that can reflect the degree of condensation;

[0077] The feature selection algorithm is used to screen and optimize the obtained feature parameters, obtain the optimal feature subset related to the condensation degree, and construct the condensation degree prediction feature set;

[0078] The obtained condensation degree prediction feature set is used to train the support vector regression algorithm to obtain a condensation degree prediction model;

[0079] The obtained condensation degree prediction model is used to predict the condensation degree of the target digital substation;

[0080] In the specific implementation, MATLAB and other tools are used to perform curve fitting and regression analysis on the experimental data to obtain a mathematical model between the condensation degree, the electromagnetic wave frequency band, and the transmission characteristic parameters, and quantitatively describe the correlation between them, such as the condensation degree is negatively correlated with the electromagnetic wave frequency and positively correlated with the transmission loss. According to the correlation law model, the optimal monitoring frequency band set that is most sensitive to condensation and least affected by the environment is determined to provide guidance for subsequent data collection. The multi-band electromagnetic wave transmission data acquisition technology is used to synchronously collect electromagnetic wave transmission data for each frequency band in the selected monitoring frequency band set to obtain a multi-band electromagnetic wave transmission data set reflecting the condensation state. Using signal processing algorithms such as wavelet transform and Fourier transform, the pre-processed multi-band electromagnetic wave transmission data is analyzed in the time and frequency domain to extract characteristic parameters such as attenuation characteristics, phase characteristics, and spectrum characteristics that can reflect the condensation degree. The extracted multiple feature parameters are screened and optimized by a feature selection algorithm based on recursive feature elimination, redundant and irrelevant features are removed, and the optimal feature subset highly correlated with the condensation degree is obtained, and a feature set for condensation degree prediction is constructed. Select machine learning algorithms such as support vector regression, take the optimal feature subset as input, and use the condensation degree as output to build a condensation degree prediction model. Use grid search and other methods to optimize the hyperparameters of SVR, such as kernel function, penalty coefficient, error tolerance, etc., to improve the fitting ability of the model. The monitoring data set is randomly divided into training set and test set according to a certain ratio, and the SVR model is trained on the training set to obtain the nonlinear mapping relationship between the condensation degree and the characteristic parameters; the performance is evaluated on the test set, and the generalization ability of the model is verified by indicators such as mean square error and determination coefficient, and the model is fine-tuned and optimized, and finally a condensation degree prediction model with superior performance is obtained. The trained condensation degree prediction model is integrated into the substation condensation monitoring system. By real-time collection of multi-band electromagnetic wave transmission data and extraction of characteristic parameters, the prediction model is input for online evaluation of condensation degree, and the condensation risk in the substation is monitored in real time. When the predicted condensation degree exceeds the preset threshold, the early warning mechanism is triggered, and an alarm is issued to the on-duty personnel through sound and light alarms, SMS notifications, etc. At the same time, condensation suppression devices such as heating and ventilation are activated to timely alleviate the condensation risk and ensure the safe operation of the substation. The condensation monitoring data and early warning records are stored in the database to form condensation big data, which provides data support for subsequent condensation law analysis and prediction model optimization, and realizes the intelligent and digital condensation monitoring of substations;

[0081] In the construction of the condensation simulation experimental platform, the following equipment and parameters can be used: the temperature range of the environmental simulation cabin is -20℃50℃, the humidity range is 30%100%, and the pressure range is 0.51.5 atmospheres; the frequency range of the multi-band electromagnetic wave transmitting module covers 1GHz1000GHz, the transmitting power is adjustable, and the maximum power is not less than 1W; the receiving sensitivity of the multi-band electromagnetic wave receiving module is better than -100dBm, the receiving frequency is synchronized with the transmitting frequency, and broadband reception can be achieved through the difference frequency technology; the sampling rate of the data acquisition module is not less than 10GSPS, the bit depth is not less than 12bit, and real-time data transmission and storage functions are supported. By controlling the temperature and humidity parameters of the environmental simulation cabin, simulating different degrees of condensation, and changing the transmitting frequency of the electromagnetic wave transmitting module, the transmission characteristics of electromagnetic waves in different frequency bands are tested. Through statistical analysis, key indicators such as transmission attenuation and phase delay are extracted to preliminarily determine the correlation between the electromagnetic wave frequency band and the condensation response. Through mathematical methods such as multivariate regression and principal component analysis, a quantitative relationship model between frequency bands, environmental parameters and transmission characteristic indicators is established, and the best monitoring frequency band is finally selected. In this process, the selection of environmental parameters, the collection and analysis of experimental data are key. You can refer to the technical settings in related condensation experiments and electromagnetic wave transmission research. For the preprocessing of multi-band electromagnetic wave transmission data, the wavelet transform combined with EMD empirical mode decomposition can be used. First, the original data is decomposed at multiple scales using wavelet transform to obtain wavelet coefficients at different frequency scales. At each scale, the wavelet coefficients are decomposed into several intrinsic mode functions and a residual component using the EMD method. According to the frequency characteristics and energy distribution of IMF, the effective IMF reflecting condensation information and the invalid IMF representing noise interference are distinguished. The invalid IMF is eliminated, the valid IMF is retained and the denoised wavelet coefficients are reconstructed. Finally, the preprocessed electromagnetic wave transmission data is obtained using the inverse wavelet transform. On this basis, the Hilbert transform can be used to calculate the instantaneous frequency and instantaneous amplitude, or the singular value decomposition can be used to extract the main eigenvectors of the data as input for subsequent feature index extraction. In the construction of the condensation degree prediction model, the kernel function of the SVR model can use the Gaussian kernel function, that is Where x and y are input feature vectors, and σ is the kernel function width parameter. The penalty coefficient C can be adjusted within the range of 0.1100, and the error tolerance ε can be adjusted within the range of 0.011. During the grid search process, a rough value grid can be determined first, such as C in {0.1, 1, 10, 100}, σ in {0.01, 0.1, 1, 10}, and ε in {0.01, 0.05, 0.1, 0.5}. The model performance under each parameter combination is determined by 5-fold cross validation, and the optimal parameter range is screened out. Then, a fine value grid is constructed near the optimal parameters, such as C in {5, 7, 9, 11, 13}, and σ in {0.5, 0.7, 0.9, 1.1, 1.3}. Repeat the above process to find the optimal parameter combination. On this basis, the learning curve can also be used to diagnose whether the model has overfitting or underfitting problems, and improve it through regularization terms, sample size and other methods. Model performance evaluation indicators can be selected from mean square error MSE, mean absolute error MAE, and determination coefficient R 2 Etc., through the comprehensive evaluation of the model's fitting effect and prediction ability through multiple indicators. In the condensation warning and mitigation mechanism, multi-level warning thresholds can be set, such as triggering a yellow warning when the condensation degree is 30%, triggering an orange warning when it is 50%, and triggering a red warning when it is 70%. Different warning levels correspond to different mitigation measures, such as starting low-power heating and natural ventilation during a yellow warning, starting medium-power heating and mechanical ventilation during an orange warning, and starting high-power heating and forced convection during a red warning. At the same time, warning messages of different frequencies and contents are sent to the on-duty personnel according to the warning level, such as once per hour, once every 30 minutes, once every 10 minutes, etc. In actual applications, the warning thresholds and mitigation measures can also be optimized according to factors such as the environmental characteristics of the substation and the equipment layout to improve the sensitivity of the warning and the effectiveness of the mitigation. For the data of multiple monitoring points, prediction models and warning mechanisms can be established respectively, and comprehensive warnings can be achieved through methods such as data fusion and logical judgment to improve the reliability of the system;

[0082] S4. Acquire the monitoring data of the target digital substation in real time, and select the electromagnetic wave monitoring frequency band based on the current environment; specifically, the steps include:

[0083] Real-time acquisition of environmental monitoring data of the target digital substation; the environmental monitoring data includes temperature data, humidity data, pressure data and light intensity data of the UB Aokuo monitoring point;

[0084] Clustering algorithm is used to perform clustering analysis on the acquired environmental monitoring data, and the target digital substation is divided into several environmental areas according to the clustering results;

[0085] Install condensation sensors in various environmental areas to monitor condensation data on the equipment surface in real time;

[0086] Select a monitoring frequency band, and use the selected monitoring frequency band to transmit condensation data;

[0087] In the specific implementation, environmental monitoring points are arranged in different areas inside the substation to collect environmental parameter data such as temperature, humidity, pressure and light intensity at each monitoring point, and the collected data is transmitted to the data processing center of the monitoring system in real time. In the data processing center, the collected environmental parameter data is preprocessed, and outliers and invalid values ​​are removed through data cleaning, data normalization and other operations, and the data is converted into a unified dimension and scale to form a standardized environmental parameter data set. According to the environmental characteristics of the substation, the range of candidate frequency bands is determined, such as extremely low frequency, ultra-low frequency, high frequency, microwave, etc. The electromagnetic wave transmission characteristics of each candidate frequency band are tested under different environmental conditions to obtain index data such as transmission attenuation, phase delay, and signal-to-noise ratio. The obtained index data is analyzed to calculate the comprehensive performance score of each frequency band under different environmental conditions. The higher the score, the less environmental interference the frequency band is. According to the comprehensive performance score, several optimal frequency bands are selected for each environmental condition, and a frequency band-environmental condition mapping table is established to form an electromagnetic wave monitoring frequency band library. The frequency band library can be established offline through field testing, simulation calculation, etc., or it can be optimized in real time through online learning, dynamic update, etc. Clustering algorithms such as K-means and DBSCAN are used to perform cluster analysis on the standardized environmental parameter data set. According to the clustering results, the internal environment of the substation is divided into several environmental areas. The environmental conditions in each area are relatively uniform, and there are significant differences in the environmental conditions between different areas. For each environmental area, according to the characteristics of its environmental conditions, several candidate frequency bands that are least affected by the environmental conditions are selected from the electromagnetic wave monitoring frequency band library to form the preferred monitoring frequency band set for the area. Condensation sensors are installed at each monitoring point. Through capacitive, resistive, optical and other sensing principles, the condensation amount on the surface of the equipment is detected in real time, and the detection value is converted into standardized condensation degree data. At the same time, the electromagnetic wave monitoring frequency band currently used by each monitoring point is recorded, and the condensation degree data is associated with the frequency band information to form a frequency band-condensation data pair. In the data processing center of the monitoring system, the collected frequency band-condensation data pairs are summarized and stored to construct a frequency band-condensation mapping data set. Parameters such as the type, installation location, and sampling frequency of the condensation sensor need to be optimized according to the characteristics of the equipment and monitoring requirements to improve the timeliness and reliability of the condensation monitoring data. When the environmental monitoring point collects new environmental parameter data, the environmental area where each monitoring point is currently located is determined through data processing and feature extraction, and the optimal monitoring frequency band that is least affected by the current environmental conditions is dynamically selected from the preferred monitoring frequency band set in the area for condensation monitoring at the monitoring point. The frequency band-condensation mapping data set is grouped according to environmental conditions and monitoring frequency bands, and the statistics such as the mean, variance, maximum, and minimum of the condensation degree in each group are calculated to evaluate the condensation monitoring performance of different frequency bands under different environmental conditions.Through statistical methods such as hypothesis testing and variance analysis, determine whether there are significant differences in condensation monitoring performance between different frequency bands, and analyze the reasons for the performance differences. According to the results of performance evaluation and difference analysis, dynamically adjust and optimize the original frequency band selection strategy, such as eliminating frequency bands with poor performance, adding new frequency bands with excellent performance, adjusting the priority of each frequency band, etc., and apply the optimized frequency band selection strategy to the dynamic frequency band selection process to achieve adaptive optimization of the monitoring frequency band. Parameters such as the statistical analysis cycle, indicator threshold, and update trigger conditions need to be flexibly set according to the real-time and accuracy requirements of condensation monitoring to balance system performance and computing costs. At the same time, make full use of machine learning algorithms to achieve automatic optimization and intelligent evolution of frequency band selection strategies, and continuously improve the accuracy and reliability of condensation monitoring. When arranging environmental monitoring points inside the substation, a grid-based layout can be used to divide the substation into several 20m×20m grids, and an environmental monitoring point is installed at the center of each grid. Each monitoring point is equipped with temperature, humidity, pressure and light intensity sensors. The sensor ranges are -40℃~80℃, 0%~100%, 500hPa~1100hPa and 0lux~100000lux, respectively, with resolutions of 0.1℃, 0.1%, 1hPa and 1lux, respectively, and the sampling frequency is 1 time / minute. The environmental monitoring point is connected to the data processing center through the RS485 bus, and the ModbusRTU protocol is used for data transmission. Each data frame includes fields such as monitoring point ID, sensor type, data timestamp and measurement value, and the communication baud rate is 9600bps. The data processing center uses the FP-growth algorithm to mine the frequent patterns of the received environmental parameter data, discover the association rules between the parameters, and use the box-whisker plot and Z-score method to detect anomalies and denoise the data to generate a standardized environmental parameter data set. For the test of electromagnetic wave transmission characteristics, an omnidirectional monopole antenna can be selected as the transmitting antenna and a logarithmic periodic antenna as the receiving antenna. The antenna gains are 2dBi and 10dBi respectively, and the standing wave ratio is less than 1.5. The spectrum characteristics and S parameters of the transmission signal are measured using a spectrum analyzer and a vector network analyzer respectively. The scanning frequency range is 10kHz to 10GHz, the number of scanning points is 1601, the intermediate frequency bandwidth is 10kHz, and the dynamic range is greater than 120dB. The antennas are placed in different environmental areas of the substation, and the antenna spacing, incident angle and other parameters are changed to measure the transmission attenuation, phase delay and signal-to-noise ratio of electromagnetic waves in different frequency bands. Each frequency band is tested 10 times and the average value is taken. The weighted summation method is used to comprehensively score the transmission characteristic indicators of each frequency band. The weight coefficient can be determined by the AHP hierarchical analysis method to obtain the comprehensive performance score of each frequency band under different environmental conditions, and the top 5 frequency bands with the highest scores are selected as the optimal monitoring frequency bands under the environmental conditions.When clustering the environmental parameter data, the K-means algorithm can be used, the number of clusters K is set to 5, the distance metric uses the Euclidean distance, the initial cluster center is selected by the K-means++ algorithm, and the number of iterations is 100. The clustering results are evaluated by the Silhouette index. If the index is greater than 0.7, the clustering effect is considered to be good, and the substation environment can be divided into 5 areas. In each area, the three frequency bands with the highest comprehensive performance scores are selected as the preferred monitoring frequency bands for the area, and the frequency band-environmental area matrix is ​​constructed. When collecting condensation data, a capacitive condensation sensor can be selected with a measurement range of 0-100%, a resolution of 0.1%, and a response time of less than 10s. Condensation sensors are arranged on the surfaces of key equipment such as insulators, busbars, and transformers in the substation. The distance between sensors does not exceed 1m, and the sampling frequency is 1 time / 10 minutes. The condensation monitoring data and the frequency band information are aligned by timestamp and stored as a frequency band-condensation time series data set. When performing statistical analysis on the frequency band-condensation data set, the following methods can be used for hypothesis testing: One-way ANOVA, significance level α = 0.5, F test to determine whether there is a significant difference in the mean condensation values ​​of different frequency bands; TukeyHSD post hoc test to determine whether the difference in the mean condensation values ​​between frequency bands is significant, and the frequency bands are clustered and sorted. The ARIMA model is used to predict the trend of the condensation time series, with a prediction step of 1 day. The autocorrelation coefficient and partial autocorrelation coefficient of the residual series are used to determine its stationarity through the LB test. The prediction results are compared with historical data, and the MAPE and RMSE indicators are calculated to evaluate the condensation monitoring performance and prediction accuracy ranking of each frequency band. The condensation prediction model is updated every day, and when the performance ranking changes, the update of the frequency band selection strategy is triggered. The frequency band selection strategy is optimized by using an improved particle swarm algorithm. The fitness function includes indicators such as condensation detection accuracy, energy consumption, and cost. The number of particles is 50, the maximum number of iterations is 200, the learning factor c1=c2=2, and the inertia weight w decays from 0.9 to 0.4 according to the cosine function. The optimized frequency band selection strategy is sent to each monitoring point, and the monitoring frequency band is dynamically updated.

[0088] S5. According to the electromagnetic wave monitoring frequency band obtained in step S4, feature extraction and analysis are performed to construct a condensation monitoring model based on electromagnetic wave transmission characteristics; specifically comprising the following steps:

[0089] Based on the maximum relevance minimum redundancy algorithm and the Relief-F algorithm, feature selection is performed on the electromagnetic wave attenuation feature data, and the attenuation feature with the greatest correlation with the condensation degree is screened out as the candidate feature set.

[0090] Feature extraction and fusion are performed on the candidate feature set. The principal component analysis method is used to extract the main component features of the attenuation feature. The independent component analysis method is used to extract the independent component features of the attenuation feature. The autoencoder is used to extract the deep nonlinear features of the attenuation feature to obtain the attenuation feature parameters.

[0091] Acquire condensation image data and video data on the device surface;

[0092] Extract image feature parameters of the acquired image data and video data, score the condensation degree according to the set requirements, and obtain a condensation observation sample data set;

[0093] The condensation observation sample data set is used to train the gradient boosting regression tree model to obtain the condensation monitoring model.

[0094] Obtain the weather forecast data, internal environment monitoring data and electromagnetic wave attenuation data of the target digital substation, and use the obtained data to train the long short-term memory neural network to obtain the condensation trend prediction model;

[0095] Use the condensation trend prediction model to predict the condensation trend of the target substation;

[0096] In the specific implementation, the electromagnetic wave attenuation data of the selected frequency band under different condensation environments are collected through the electromagnetic wave transmission test system. The attenuation data includes time series data of multiple dimensions such as amplitude attenuation, phase delay, and spectrum change. The sampling frequency is not less than 10kHz, and the number of samples collected in each frequency band is not less than 10,000. The collected electromagnetic wave attenuation data is preprocessed, the data is denoised and smoothed by wavelet transform, the data is scaled and distributed by Z-score normalization, and the missing data is repaired by interpolation to form a standardized attenuation data set. The MRMR maximum relevance minimum redundancy algorithm and the Relief-F algorithm are used to select features for the standardized attenuation data set, and the 10 attenuation features with the greatest correlation with the condensation degree are selected as candidate feature sets. The feature selection process ensures the stability and reliability of the results through 10-fold cross validation. The candidate feature set is further extracted and fused, and the principal component analysis PCA is used to extract the main component features of the attenuation data, the independent component analysis ICA is used to extract the independent component features of the attenuation data, and the autoencoder is used to extract the deep nonlinear features of the attenuation data, and finally a set of comprehensive attenuation feature parameters are obtained. Several key locations where condensation is common in the substation, such as insulators, busbars, transformers, etc., are selected as condensation observation points. High-definition cameras and visual sensors are installed at each observation point to collect images and video data on the equipment surface in real time. Computer vision algorithms are used to analyze image and video data, extract characteristic parameters such as the number of pixels and grayscale value in the condensation area, and obtain physical quantity indicators such as condensation amount and coverage area. Based on the condensation physical quantity indicators, combined with factors such as the insulation performance of electrical equipment and meteorological conditions, the condensation degree at each sampling moment is scored. The score ranges from 0 to 10 points. The higher the score, the more serious the condensation degree. The physical quantity indicators and scores are matched one by one to generate a condensation observation sample data set. Each sample includes fields such as collection time, collection location, physical quantity indicators, and condensation degree score. The gradient boosting regression tree (GBRT) algorithm is used to train the condensation monitoring model with attenuation feature parameters as input and condensation degree score as output. The model's hyperparameters, such as the number of trees, depth, and learning rate, are optimized through grid search. The model performance is evaluated through 5-fold cross validation. According to the actual application requirements, the evaluation criteria of the model performance are reasonably set, and the optimal model is selected for deployment and application. The trained condensation monitoring model is integrated into the substation condensation online monitoring system, and the electromagnetic wave attenuation data of the selected frequency band is collected in real time. The attenuation feature parameters are extracted and input into the model to continuously predict the condensation degree score. The meteorological forecast data of the substation area, including meteorological elements such as temperature, humidity, pressure, wind speed, and rainfall in the future period, are collected as external input features for condensation prediction. The meteorological forecast data is aligned and fused with the environmental monitoring data and electromagnetic wave attenuation data inside the substation in time and space to form a multi-source heterogeneous condensation prediction sample set.Using time series prediction models such as long short-term memory (LSTM) neural networks, with historical condensation levels and multi-source physical quantities as input and future condensation levels as output, a multi-step, multi-variable condensation trend prediction model is trained. The input data of the model is updated in real time using a sliding window method to dynamically predict the trend of condensation levels in the future, and the alarm threshold and outage threshold are adjusted in a timely manner according to the prediction results. The condensation level prediction results are compared with the preset thresholds. If the warning threshold is exceeded, an audible and visual alarm is triggered to alert the operation and maintenance personnel. If the outage threshold is exceeded, the equipment is forced to shut down for protection. The condensation trend prediction results are published to relevant business systems and decision-makers through Web services to assist them in formulating condensation prevention measures and operation and maintenance strategies. In practical applications, attention should be paid to key technical issues such as spatiotemporal data fusion, multi-step prediction, and model update to improve the timeliness and accuracy of predictions. At the same time, the visualization of prediction results and human-computer interaction design should be done well to improve the system's availability and user experience. When collecting electromagnetic wave attenuation data, a spectrum analyzer with a frequency range of 1MHz to 10GHz, a dynamic range greater than 120dB, and a sensitivity better than -130dBm can be selected, and a horn antenna with a gain greater than 30dBi and a standing wave ratio less than 1.2 can be used. More than 10 collection points are arranged in different areas of the substation, and each collection point collects data every 10 minutes. Each candidate frequency band is scanned 100 times during each collection, and the average value is taken as the attenuation data of the frequency band. In the data preprocessing stage, the Daubechies wavelet basis function is used to perform a 6-level discrete wavelet transform on the attenuation data, and then the Donoho threshold is used to perform soft threshold denoising on the wavelet coefficients, and then wavelet reconstruction is performed to obtain the denoised attenuation data, and the Z-score method is used to normalize the data. For data with a missing rate of less than 10%, linear interpolation is used to fill in, and data with a missing rate of more than 10% is directly eliminated. In the feature selection stage, the MRMR algorithm is used to calculate the mutual information between each attenuation feature and the degree of condensation, and the 10 features with the largest mutual information are selected as the candidate feature set. Then, the Relief-F algorithm is used to calculate the correlation statistics of each candidate feature, and the candidate features are sorted according to the correlation statistics. The top 5 features are selected as the final feature selection results. In the feature extraction stage, the PCA algorithm is used to reduce the dimension of the candidate feature set, and the principal component with a cumulative contribution rate greater than 95% is selected as the main component feature. The FastICA algorithm is then used to extract the independent components of the candidate feature set, and the independent components with a kurtosis greater than 5 are selected as the independent component features. Finally, the autoencoder is used to perform nonlinear transformation on the candidate feature set, and the hidden node output with a reconstruction error less than 0.01 is selected as the deep feature.In the process of condensation observation data collection, a high-definition digital camera with a resolution greater than 10 million pixels and a laser condensation sensor with a measurement range of 0-100% and a resolution better than 0.1% can be selected. Two cameras and four sensors are installed at key locations such as insulators, busbars, and transformers in the substation. Images and condensation data are collected synchronously every 30 minutes. The image data is preprocessed and input into the faster R-CNN target detection algorithm to extract the location and size information of the condensation area. The condensation data is used as the condensation parameter after outlier removal and smoothing filtering. The image parameters and condensation parameters at the same time are combined into an observation sample. In the process of condensation degree scoring, each condensation observation sample is scored independently. According to the condensation image and condensation amount data in the sample, combined with the equipment layout, meteorological conditions, operating conditions and other factors of the substation, a score of 0 to 10 points is given. The median of the 10 scoring results is taken as the final score of the sample. Then the image parameters, condensation amount parameters and scoring results of all samples are summarized into a condensation observation sample data set. The data set is randomly divided into a training set and a test set in a ratio of 8:2. In the condensation monitoring model training stage, the GBRT algorithm is used to fit the training set data, and the model hyperparameters are optimized by grid search. The parameter search range is: the number of trees [50, 100, 200, 500], the maximum depth of the tree [3, 5, 7, 9], the learning rate [0.01, 0.05, 0.1, 0.2], the loss function is the mean square error, and the evaluation index is the determination coefficient R. 2 Then the model was evaluated using the 5-fold cross validation method and the average R 2The set of hyperparameters with the largest value is taken as the optimal parameter, and finally a trained GBRT condensation monitoring model is obtained. In the model application stage, the trained GBRT model is used to predict the newly collected electromagnetic wave attenuation data. First, the characteristic parameters of the attenuation data are extracted, and then input into the model to obtain the condensation degree score. The score is then compared with the pre-set threshold. If the score exceeds the warning threshold, a warning SMS is sent to the on-duty personnel. If the score still exceeds the shutdown threshold after three consecutive warnings, the equipment is immediately triggered to be forced to shut down. At the same time, a multivariate time series prediction model is constructed using the LSTM network. The input variables include the condensation degree score, electromagnetic wave attenuation characteristics, meteorological elements, etc. in the past 24 hours. The output variables are the condensation degree scores for the next 1 hour, 2 hours, and 4 hours. The input data of the model is updated every 1 hour and a prediction is made. The prediction results are used to assist in adjusting the warning and shutdown thresholds. The main parameters of the LSTM model are set as follows: input time step 24, number of hidden layer neurons 128, number of output neurons 3, learning rate 0.001, batch size 32, number of training rounds 100, activation function tanh, loss function mean square error, optimization algorithm Adam. When deploying the application, use the FlaskWeb framework to build a RESTfulAPI service, receive real-time data from monitoring equipment, call the prediction model for calculation, and push the prediction results to the Web front end in real time through the WebSocket protocol. The front end uses the Echarts component to visualize the prediction results, and sets up human-computer interaction functions to allow users to adjust threshold settings and query historical data;

[0097] S6. Optimizing the condensation monitoring model constructed in step S5 based on the fused environmental parameters and electromagnetic wave parameters; specifically comprising the following steps:

[0098] The acquired environmental parameters and electromagnetic wave parameters are preprocessed through data alignment, time series interpolation and outlier removal algorithms to obtain fused feature vectors;

[0099] The condensation monitoring model constructed in step S5 is optimized by using the fused feature vector;

[0100] In the specific implementation, environmental sensors and electromagnetic wave sensors are deployed in different areas inside the substation to collect environmental parameters such as temperature, humidity, pressure, and characteristic parameters such as electromagnetic wave amplitude, phase, and spectrum. The sampling frequency of the sensor is not less than 1 time / minute, and the range covers the actual working range of the substation. The data resolution is better than 0.1 level, and the collected environmental parameters and electromagnetic wave characteristic parameters are uploaded to the edge computing gateway in real time through the Internet of Things communication protocol such as MQTT. In the edge computing gateway, the collected environmental parameters and electromagnetic wave characteristic parameters are fused and processed, and the parameter data from different sources and different spatiotemporal granularity are integrated into a unified time series data set to form a fused feature vector containing environmental characteristics and electromagnetic wave characteristics, and the fused feature vector is encoded using the UUID algorithm to generate a globally unique feature identification code. The local fused feature data set is uploaded to the cloud big data platform through a secure communication tunnel, and the massive fused feature data is stored, cleaned, indexed, and calculated using big data processing frameworks such as Hadoop and Spark. Specifically, the fused feature data is stored in the HDFS distributed file system in Parquet columnar storage format, and the compression and encoding mechanism of Parquet is used to improve the storage efficiency and reading performance of the data; SparkSQL is used to perform ETL processing on the fused feature data, and data cleaning and conversion are achieved through operations such as data filtering, data conversion, and data aggregation, and the query optimizer and memory computing engine of SparkSQL are used to accelerate the speed of data processing; the SparkMLlib machine learning library is used to perform feature engineering and statistical analysis on the fused feature data, such as feature selection, feature scaling, and correlation analysis, to mine the inherent laws and correlations of the data; SparkStreaming is used to process the newly collected fused feature data in real time, and the real-time cleaning and calculation of the data are achieved by setting sliding windows and trigger intervals, and the processing results are written to distributed memory databases such as Redis for subsequent model training and prediction. The multi-dimensional display and interactive analysis of fused features are realized through data visualization technology, and the correlation laws and abnormal patterns between environmental parameters and electromagnetic wave feature parameters are discovered. Based on the cloud-based fusion feature data set, a machine learning sample set for condensation monitoring is constructed. The sample features include fusion feature vectors, feature identification codes, condensation degree, etc., and are divided into training sets and test sets in a ratio of 8:2. Gradient boosting decision tree algorithms such as XGBoost and LightGBM are used to train the condensation monitoring model with training set samples as input.In the process of model training, data enhancement and transfer learning mechanisms are added. By performing data enhancement operations such as random rotation, translation, and noise addition on training samples, the model's adaptability to environmental changes is improved. Through transfer learning technology, the condensation monitoring model trained in other substation scenarios is used as a pre-training model, and fine-tuned and retrained in the new substation environment to reduce the demand for training samples, accelerate the convergence of the model, and improve the robustness of the model. The feature expression ability and generalization performance of the model are further improved through feature engineering and parameter optimization. The model evaluation indicators include accuracy, recall rate, F1 value, etc., which comprehensively weigh the model's fitting ability and prediction ability. The condensation monitoring model is robustly analyzed and the generalization performance is evaluated. Through repeated experiments, the statistical indicators such as the mean, variance, and confidence interval of the model output results are evaluated to judge the model's anti-interference ability and prediction reliability. The classification performance of the model is evaluated using ROC curves and confusion matrices, and the model structure and parameters are optimized in a targeted manner. The trained and optimized condensation monitoring model is deployed to the edge computing gateway to achieve localized real-time monitoring and early warning. When the edge gateway collects new environmental parameters and electromagnetic wave characteristic parameters, it automatically performs data fusion and feature extraction, inputs the fused features into the condensation monitoring model, calculates the condensation degree prediction value online, triggers the corresponding level of condensation warning based on the comparison result between the prediction value and the preset threshold, and pushes the warning information to the on-duty personnel in a timely manner through SMS, email, voice call, etc. At the same time, the monitoring data, prediction results, warning status and other information are transmitted back to the cloud to achieve closed-loop optimization of cloud-edge collaboration. Build a data analysis and model optimization platform on the cloud to receive feedback information such as monitoring data, prediction results, warning status, etc. from the edge, and store it in the time series database and object storage to form a historical data set. Use time series data analysis and visualization tools to perform trend analysis, anomaly detection, pattern discovery, etc. on historical data to explore the laws and problems behind the data. Use AutoML tools to automatically optimize the condensation monitoring model, find the optimal model structure and parameter configuration through hyperparameter search, model integration, automatic feature engineering and other technologies, and use online learning algorithms to incrementally train and update the model to adapt to the changing data distribution and environmental conditions. The optimized model and configuration parameters are sent to the edge, and the monitoring model and warning threshold of the edge are updated to achieve online update and performance improvement of the model. The effect and benefits of the new model are evaluated through experimental mechanisms such as A / B testing and multi-armed bandit, forming a data-driven continuous optimization closed loop. In the edge gateway, the environmental parameter data is time-aligned using a linear interpolation algorithm, and the electromagnetic wave characteristic parameter data is removed using the Tukey algorithm. Then, the two types of data are normalized using a standard method and converted into standard normal distribution data with a mean of 0 and a variance of 1.The environmental parameter data and electromagnetic wave characteristic parameter data are merged according to the time window of 1 minute by timestamp to generate fused feature vectors. Each vector contains 10 environmental features and 50 electromagnetic wave features. Then, the UUID algorithm is used to generate a 128-bit feature identification code to form the final fused feature data set. The QoS level of the MQTT message is set to 2 to ensure the reliability of data transmission. The fused feature data set is stored in the Parquet format on HDFS, partitioned according to the acquisition time and feature identification code, and the Parquet file is compressed using the Snappy algorithm. SparkSQL is used to process the fused feature data, and statistical indicators such as mean, variance, and correlation coefficient of environmental parameters and electromagnetic wave parameters are generated by filtering abnormal data, selecting key features, and agg function data aggregation. The random forest algorithm is used to evaluate the feature importance of the fused feature data, calculate the Gini index of each feature, and select the top 20% features of importance as model input. The Pearson correlation coefficient, Spearman rank correlation coefficient and other methods are used to analyze the correlation between environmental parameters and electromagnetic wave parameters, and the analysis results are displayed in a visual way such as heat map and scatter plot. Based on the Word2Vec algorithm of SparkMLlib, the fused feature vector is converted into a 256-dimensional word vector for subsequent model training and similarity calculation. Using the sliding window mechanism, the window length is set to 10 minutes, the sliding step is set to 5 minutes, the real-time data stream is accumulated and counted, and anomaly detection is performed, and the results are written to the Redis cluster. The data expiration time is set to 24 hours. The training sample set of the condensation monitoring model contains 10,000 positive samples and 50,000 negative samples. The positive samples are the fused feature vectors when condensation occurs, and the negative samples are the fused feature vectors when condensation does not occur. The samples are arranged in chronological order, with the first 80% as the training set and the last 20% as the test set. The XGBoost algorithm is used to train the binary classification model, gbtree is used as the base classifier, the maximum depth of the tree is set to 6, the learning rate is 0.1, the regularization coefficient is 1, the objective function is the cross entropy loss function, and the evaluation indicators are AUC value, F1 score, precision and recall. The grid search method is used to tune the hyperparameters such as the number of trees and the feature sampling ratio. Each set of hyperparameter combinations is trained for 10 rounds. The average value of the AUC value is taken as the scoring criterion, and the hyperparameter combination with the highest average AUC value is selected as the optimal parameter. The training samples are subjected to data enhancement operations such as 5-degree random rotation, 10% random scaling, and 0.01 Gaussian noise. The enhanced samples are added to the training set to improve the generalization ability of the model. The condensation monitoring model trained in other 10 substations is used as a pre-training model. 800 groups of samples are used to fine-tune the model and optimize the weight parameters of the fully connected layer. Transfer learning can improve the generalization performance of the model by 3 to 5 percentage points.The condensation monitoring model was cross-validated 10 times, with the training set and validation set randomly divided each time. The model was trained on the training set and evaluated on the validation set to obtain the evaluation indicators of the 10 models. The mean and standard deviation of the indicators were calculated. The mean reflects the average performance of the model, and the standard deviation reflects the stability of the model. The AUC mean reached above 0.95 and the standard deviation was less than 0.01, indicating that the model performance was robust. Using the Monte Carlo simulation method, 10,000 sets of data were randomly generated for the environmental parameters, electromagnetic wave parameters, condensation labels, etc. of the model input. The data distribution was the same as the real distribution. The performance of the model on random data was evaluated. If the generalization performance did not drop by more than 5%, it indicated that the model had good robustness. The model was deployed on the edge gateway and the FlinkML library was used to load and predict the model. The average prediction delay was less than 50ms and the resource utilization rate did not exceed 50%, which could meet the performance requirements of real-time warning. When the probability of condensation exceeds 70%, the first-level warning is triggered, and a warning message is sent to the on-duty personnel, and the equipment heating and ventilation are controlled at the same time. When the probability of condensation exceeds 90%, the second-level warning is triggered, and a voice call is sent to the operation and maintenance person in charge, and the equipment is controlled to be shut down and isolated. After each warning is triggered, the warning data is uploaded to the cloud, and the warning status and processing records in the cloud are updated. InfluxDB time series database is used in the cloud to store condensation monitoring and warning data, with a data retention policy of 1 year and a data sampling granularity of 1 minute. Grafana is used to display the time series curve of condensation, the spatiotemporal distribution of warnings, and statistical charts of environmental parameters and electromagnetic wave parameters. The Grafana alarm plug-in is configured. When the edge data delay exceeds 5 minutes and the number of warnings exceeds 10 times, an alarm email is sent to the system administrator. Using the Auto-Sklearn automatic machine learning tool, with historical data as input and model performance as the optimization goal, a series of modeling steps such as data preprocessing, feature engineering, algorithm selection, and hyperparameter tuning are automatically searched to obtain an optimal model Pipeline, and the model is exported in ONNX format and deployed on the edge for prediction. Use TensorflowServing to build the model service, use Flask to implement the Web service interface, receive prediction requests from the edge, return prediction results, and control the average service delay within 100ms. Use the A / B test method to randomly divert edge requests at a ratio of 9:1, request the current production model and the newly trained model respectively, record the prediction results and prediction time of the two models, and use the t-test method to compare whether the performance difference between the two models is significant. When the performance of the new model is significantly better than the current model and the prediction time is not more than 1.5 times that of the current model, start the model replacement process, deploy the new model online, and update the configuration parameters of the edge.Use Prometheus and Grafana to build a cloud-edge monitoring platform, collect and summarize various monitoring indicators of edge nodes in the cloud, including CPU utilization, memory utilization, disk IO, network bandwidth, data latency, model prediction time, etc. Use the LSTM algorithm to predict the trend of monitoring indicators, and use the isolation forest algorithm for anomaly detection. When the indicator exceeds the threshold or the predicted trend is abnormal, the cloud alarm is triggered to notify the on-duty personnel to check and handle it, ensuring the high reliability and high availability of the condensation monitoring service.

[0101] S7. Based on the optimized condensation monitoring model obtained in step S6, a condensation risk prediction result of the target digital substation is generated; specifically, the steps include:

[0102] The optimized condensation monitoring model is used to predict the condensation risk of the target digital substation;

[0103] According to the predicted condensation risk probability, clustering algorithm is used to perform clustering and grouping, and different risk levels are marked;

[0104] The monitoring grid of the target digital substation is set as an undirected weighted graph, where each monitoring point corresponds to a node of the graph. There is a weighted edge between adjacent monitoring points, and the weight of the edge is set; the Node2Vec algorithm is used to optimize the node collinearity probability to obtain the low-dimensional embedding vector of the node; the IDW algorithm is used to interpolate and predict the risk level of unknown nodes; based on the interpolation prediction results, the substation geographic information map is used as the base map and the monitoring points are used as control points to generate a risk distribution map; areas with different risk levels are marked;

[0105] In the specific implementation, the monitoring grid is divided inside the substation. According to the actual area and equipment distribution of the substation, the grid size is set to 10m×10m or 20m×20m. A condensation monitoring point is arranged in each grid to monitor the condensation status of the electrical equipment in the grid. The data collected by each monitoring point is transmitted back to the condensation monitoring cloud platform in real time through the wireless sensor network, and the HDFS distributed file system and HBase distributed database are used in the cloud platform to store and manage the monitoring data. The MapReduce distributed computing framework is used to process the massive monitoring data transmitted back by each monitoring point in parallel, and the data quality is improved through data preprocessing operations such as data cleaning, anomaly detection, and missing value filling. The condensation monitoring data is then divided into time windows of fixed length, such as 30 minutes, 1 hour, etc. Each time window contains multiple time steps to form a two-dimensional matrix. The two-dimensional matrix is ​​converted into a grayscale image. Each time step corresponds to a column of pixels in the image, and different monitoring indicators correspond to different channels of the image. The converted image is input into the pre-trained CNN model to extract high-level features, such as pooling layer output, fully connected layer output, etc., to obtain a fixed-length feature vector. Among them, the CNN model can choose classic structures such as LeNet, AlexNet, VGGNet, or design a customized network structure according to the specific data characteristics. The Adam optimizer can be used for model training, the learning rate is set to 0.001, the batch size is set to 32, and the number of training rounds is set to 100. After each round, the accuracy and loss of the model are evaluated on the validation set, and overfitting is prevented according to the early stopping rule. The effect evaluation of the model can use the leave-one-out cross-validation method, divide the data set into n parts, select one of them as the test set each time, and the remaining n-1 parts as the training set, repeat the experiment n times, and take the average value as the final evaluation index. Finally, the feature vector is used as the input of the condensation monitoring model to judge the condensation risk probability of each monitoring point in real time. According to the level of condensation risk probability, the clustering algorithm is used to automatically group the risk probability and use different colors such as green, yellow, orange, and red to identify different risk levels, and generate a mapping relationship table between condensation risk level and color. The substation monitoring grid is abstracted as an undirected weighted graph. Each monitoring point corresponds to a node of the graph. There is a weighted edge between adjacent monitoring points. The weight of the edge can be Euclidean distance, correlation coefficient, etc. The Node2Vec algorithm is used to learn the embedding vector of the node. Node2Vec learns the low-dimensional vector representation of the node by optimizing the probability of node co-occurrence, which can capture the structural similarity and homogeneity between nodes.Then, the IDW algorithm is used to interpolate and predict the risk level of unknown nodes. IDW is a local interpolation algorithm based on spatial correlation. The main parameters of the Node2Vec algorithm include embedding dimension, walk length, walk number, etc. The grid search method can be used to tune the parameters and select the parameter combination with the best effect on the validation set. The main parameters of the IDW algorithm include the number of nearest neighbor points and distance power. The cross-validation method can be used to select the optimal parameters and evaluate the accuracy and robustness of the interpolation. Finally, the interpolation results are mapped to the grid map to obtain the continuous condensation risk distribution of the entire substation. Based on the interpolation prediction results, open source GIS tool libraries such as Leaflet are used, with the substation geographic information map as the base map and the monitoring points as the control points to generate a heat map of the substation condensation risk distribution. Areas with different risk levels are rendered with different colors and transparency. The darker the color and the lower the transparency, the higher the risk level. Contour lines, wind vanes and other graphic symbols are used to mark the changing trend of risk levels and key risk points. On the basis of the heat map, the building layout map, equipment layout map and other thematic layers inside the substation are superimposed to mark the key areas and key equipment of the substation. The interactive WebGIS system is developed using the Shiny package of R language, which provides interactive functions such as zooming in and out, panning and roaming, and layer control, allowing users to customize the threshold and color scheme of the condensation risk level, dynamically adjust the display effect of the risk distribution map, and realize the intuitive display of the condensation risk. In addition, multi-source data such as risk distribution maps, environmental monitoring data, and weather forecast data can also be cleaned and format converted, and key fields such as time, location, and indicators can be extracted and converted into a standard spatiotemporal data cube format. Pre-aggregation OLAP engines such as Apache Kylin are used to pre-calculate the data cube in multiple dimensions and multiple granularities to generate aggregate indicators and statistical reports, such as the average risk level on different time scales and the number of high-risk monitoring points in different areas. Statistical drawing libraries such as Seaborn are then used to visualize the aggregate indicators and generate various charts such as line charts, bar charts, and heat maps to reveal the correlation and trend between the risk level and environmental factors and weather conditions, providing a more comprehensive basis for condensation prevention. Among them, the design of the data cube needs to consider business needs and data characteristics, select appropriate dimensions, metrics and granularity, and balance query performance and storage costs; the association of multi-source data needs to be based on spatiotemporal attributes, such as timestamps, geographic coordinates, etc., which can be achieved using SQL operations such as JOIN and GROUP BY; correlation analysis can be measured using indicators such as Pearson correlation coefficient and mutual information, and trend analysis can be performed using non-parametric methods such as Mann-Kendall rank correlation test and Theil-Sen estimation;.

[0106] S8. According to the prediction result obtained in step S7, the optimal anti-condensation measures are given to the target digital substation to complete the condensation monitoring of the target digital substation; specifically, the steps include:

[0107] According to the prediction result obtained in step S7, the area where the condensation risk is greater than the set threshold is obtained, and the specific location and coverage are determined in combination with the layout data of the target substation to obtain a location list of high-risk areas for condensation;

[0108] Build a knowledge base of anti-condensation measures;

[0109] In the anti-condensation measures knowledge base, the environmental characteristics of the high-risk area for condensation are compared with the environmental characteristics data in the anti-condensation measures knowledge base for similarity, and the anti-condensation measures with similarity higher than the set threshold are selected as candidate measures;

[0110] Among the candidate measures, the cost data of each candidate measure is calculated according to the installation location and installation method of the candidate measure, and the changes in regional environmental parameters after the implementation of each candidate measure are calculated through the three-dimensional model of the target digital substation, and the anti-condensation effect of the candidate measures is evaluated. According to the cost data and the anti-condensation effect, the final anti-condensation measure is determined to complete the condensation monitoring of the target digital substation;

[0111] In specific implementation, the list attributes include area number, center coordinates, boundary coordinates, area, perimeter, coverage equipment, etc. According to the high-risk area location list, extract the environmental monitoring data in the area, including temperature, humidity, pressure, wind speed, etc., use historical data of more than 5 years to calculate the annual average, seasonal average, monthly average and other statistical indicators of each environmental parameter, and combine with the meteorological knowledge base to summarize the environmental characteristics of the area and form a description of the environmental characteristics of the high-risk area. The description attributes include area number, temperature characteristics, humidity characteristics, pressure characteristics, wind speed characteristics, condensation mode, etc. The measure attributes include measure number, measure name, applicable conditions, technical principles, installation location, investment cost, operation and maintenance cost, service life, etc. When calculating the similarity between the applicable conditions of the anti-condensation measures and the regional environmental characteristics, the applicable conditions of the measures and the regional environmental characteristics are represented as a multidimensional feature vector, each dimension corresponds to an environmental parameter, and the distance between the characteristic values ​​of the applicable conditions of the measures and the regional environmental characteristics is calculated for each dimension. The distance of each dimension is weighted averaged using commonly used distance metrics such as Euclidean distance and Manhattan distance to obtain a comprehensive distance. The weight is determined according to the importance of each environmental parameter on condensation. The comprehensive distance is converted into similarity through the Gaussian kernel function. By setting different parameter values, the decay speed of the similarity is controlled and the matching sensitivity is adjusted. According to actual needs and experience, appropriate parameter values ​​are selected, the similarity threshold is set, and measures with similarity greater than the threshold are taken as candidate measures. They are sorted from high to low according to the similarity to form a list of optional measures in the area. For each measure in the optional measures list, its specific installation location and installation method are determined in the 3D model of the substation according to its installation location attributes, and computational fluid dynamics software is used to simulate the temperature and humidity distribution and airflow field in the area after the implementation of the measures. In the CFD simulation, according to the actual layout and equipment parameters of the substation, the boundary conditions, initial conditions, turbulence models, etc. are reasonably set, the calculation domain is meshed, and appropriate fluid physical parameters and solution settings are selected to ensure the convergence and accuracy of the simulation results. By comparing the temperature and humidity distribution and airflow streamlines before and after the implementation of the measures, the anti-condensation effect of the measures is evaluated, and a measure effect evaluation report is formed. The report attributes include measure number, installation location, temperature effect, humidity effect, airflow effect, comprehensive effect, etc. According to the measure effect evaluation report and the cost attributes of the measures, the cost performance of each measure is calculated, and the cost performance is sorted from high to low to form a recommended list of measures. The list attributes include area number, measure name, cost performance, investment cost, operation and maintenance cost, comprehensive effect, etc. The measure with the highest cost performance is taken as the optimal measure for the area. For the dehumidification device type among the optimal measures, the operation control strategy of the dehumidification device is optimized according to the regional environmental characteristics and the technical parameters of the dehumidification device, including start and stop time, dehumidification capacity, regeneration cycle, etc.Using intelligent control algorithms such as fuzzy control and predictive control, such as fuzzy PID control and adaptive predictive control, the control model of the dehumidification device is designed, and a dynamic mapping relationship between temperature and humidity, dehumidification capacity, and energy consumption is established. Through offline simulation and online learning, the control parameters are adaptively adjusted to minimize dehumidification energy consumption and improve dehumidification efficiency. For the optimal measures that belong to the ventilation optimization type, the operation control strategy of the ventilation system is optimized according to the regional environmental characteristics and the design parameters of the ventilation system, including ventilation time, ventilation volume, fan power, etc. A ventilation system optimization model is established, and the ventilation control parameters are used as decision variables, the probability of condensation, ventilation energy consumption, equipment life, etc. are used as optimization targets, and the environmental parameter range, fan rated power, ventilation volume range, etc. are used as constraints to form a multi-objective optimization problem. Heuristic search algorithms such as particle swarm optimization and genetic algorithm are used to search for the optimal combination of ventilation control parameters. In the particle swarm optimization algorithm, the position vector of each particle is encoded as a set of ventilation control parameters, a set of particles is randomly initialized, the fitness function value of each particle is calculated, the individual optimal position and the global optimal position are updated, the speed and position of the particles are adjusted, and the iteration is repeated until convergence, and the global optimal solution is output as the optimal ventilation control strategy. The optimal anti-condensation measures in each area are formed into a decision report and reported to the substation management personnel. The report attributes include area number, condensation risk level, environmental characteristics description, optimal measure name, cost-effectiveness of measures, energy saving and emission reduction benefits, etc. After the management personnel review and approve, the optimal measures will be included in the substation condensation control program library for subsequent implementation. According to the real-time update of the condensation risk distribution map, the above process is triggered periodically to re-evaluate the risk level changes and environmental characteristics changes in each area and adjust the existing anti-condensation measures. Combined with factors such as equipment operating conditions and weather forecasts, when the condensation risk level rises by more than 5%, or the environmental parameters fluctuate by more than 10%, the optimization process is triggered, and the latest monitoring data and external conditions are used to update the anti-condensation measures to form a closed-loop control of condensation risk. Through the collaboration of the cloud big data platform and the edge intelligent gateway, the anti-condensation measures can be continuously optimized and adaptively updated to improve the timeliness and pertinence of the plan. In terms of deployment and implementation, the microservice architecture and containerization technology are used to decouple each functional module into independent microservices, and container orchestration platforms such as Kubernetes are used for elastic scheduling and management to improve the scalability and maintainability of the system. At the same time, unified data standards and interface specifications are established to achieve seamless integration and data sharing between subsystems, providing intelligent decision-making support for the entire life cycle of substation condensation management.When identifying high-risk areas, the following methods can be used to segment the condensation risk distribution map. A segmentation algorithm based on regional growth, such as a regional merging algorithm, is used. Pixels with a risk level greater than 75% are used as seed points. The risk level difference is less than 5% as the growth criterion. Adjacent high-risk pixels are merged into one region to obtain the segmentation result of the high-risk area. Then, multi-scale geometric features are extracted for each segmented region, including area, perimeter, rectangularity, circularity, etc., and regions with an area greater than 100 square meters and a rectangularity greater than 0.8 are selected as candidate regions based on the empirical threshold. Finally, the candidate regions are superimposed and analyzed with the substation geographic information map and the equipment layout map, and the location information such as the center coordinates and boundary coordinates of the region, as well as the attribute information such as the type and quantity of the main equipment covered, are extracted to form a list of high-risk area locations. When calculating the statistical indicators of environmental parameters, an incremental learning method can be used. There is no need to store massive historical data. Only a set of statistics, including the number of samples, the sum, the sum of squares, etc., need to be maintained. Whenever new monitoring data arrives, the statistics are dynamically updated using a recursive formula. For example, the recursive formula for the mean is: Where μ_n is the mean after the nth sample, μ_{n-1} is the mean after the n-1th sample, and x_n is the nth sample value. Statistics such as variance and standard deviation also have similar recursive formulas. When performing monthly, quarterly, and annual statistics, it is only necessary to use the current statistics to calculate the corresponding indicator value at the corresponding time node. When matching anti-condensation measures, there are many measures in the knowledge base and the applicable conditions are different. In order to improve the matching efficiency and accuracy, a semantic-based multi-stage filtering method can be used. First, the text attributes of the measures, such as the name of the measures, technical principles, etc., are segmented and semantically vectorized. A pre-trained word vector model such as Word2Vec is used to map each word into a real-valued vector of a fixed length. Then, the semantic vectors of the measures are clustered, and the measures with similar semantics are clustered into one category. When optimizing the control strategy of the dehumidification device, the model predictive control method can be used to establish a dynamic model of the dehumidification process to describe the nonlinear relationship between temperature and humidity, dehumidification capacity, and energy consumption, such as a partial differential equation model based on the law of conservation of energy and mass. Then, in each control cycle, according to the current temperature and humidity measurement values ​​and the temperature and humidity prediction values ​​in the future, solve an optimal control problem in a finite time domain, such as minimizing the cumulative energy consumption in the prediction time domain, constraining the dehumidification capacity not to be lower than a certain threshold, and obtaining the optimal dehumidification capacity sequence in the future. Finally, the first value of the sequence is used as the dehumidification capacity set value in the current control cycle to control the operation of the dehumidification device. In the next control cycle, repeat the above calculation to achieve rolling optimization. In the process of implementing anti-condensation measures, it is necessary to continuously monitor and evaluate the performance of the system to promptly discover and solve problems. To this end, a monitoring system based on a microservice architecture can be deployed. The system consists of multiple independent monitoring services, including data collection services, data storage services, data analysis services, and alarm services. Each service runs in an independent Docker container and communicates through REST API. The data collection service is responsible for collecting performance indicator data from each subsystem, such as the energy consumption, temperature and humidity of the dehumidification device, the wind speed and current of the ventilation system, and pushes the data to the message queue; the data storage service is responsible for persisting the data in the time series database and providing query and aggregation APIs; the data analysis service is responsible for regularly reading data from the database, calculating key performance indicators, such as energy efficiency ratio, humidity qualification rate, etc., using anomaly detection algorithms to identify anomalies, and pushing the analysis results to the message queue; the alarm service is responsible for subscribing to the message queue and automatically sending alarm notifications, such as emails and text messages, when anomalies are detected according to the preset alarm rules. Each service is decoupled through the message queue, and can be independently expanded and upgraded. The entire system has good scalability and fault tolerance. Operation and maintenance personnel can view the system's operating status and performance indicators through the Web interface to promptly discover and locate problems;

[0112] S9. Establish a condensation monitoring and protection platform to achieve condensation monitoring, data management and protection measures optimization of the target digital substation; specifically include the following steps:

[0113] Perform data preprocessing on the acquired data to generate a condensation monitoring data set;

[0114] Based on support vector machine and convolutional neural network, an initial model for condensation risk assessment is constructed, and the generated condensation monitoring data set is used for training to obtain a condensation risk assessment model.

[0115] The obtained condensation risk assessment model is used to obtain the condensation risk assessment result, and combined with the three-dimensional model of the target digital substation, a condensation risk distribution map is generated;

[0116] According to the generated condensation risk distribution map, anti-condensation measures are dynamically optimized based on a multi-objective planning scheme;

[0117] In the specific implementation, through the Internet of Things technology, the temperature and humidity sensors, dew point meters, conductivity meters and other condensation monitoring equipment in the substation are connected to the intelligent management platform to realize unified management and remote control of the equipment. The OneM2M and OMALwM2M general Internet of Things standards and protocols are adopted to improve the interoperability and scalability of the system. For different types of monitoring equipment, a secure communication protocol and a lightweight data serialization format are used to pre-process, compress and encrypt data on the device side through the edge computing gateway to reduce the computing and storage burden on the cloud. The management platform groups and encodes the equipment according to the equipment type, location, and functional attributes to form an equipment ledger, and uses the MQTT lightweight communication protocol to establish a real-time two-way communication link with the equipment to regularly obtain the equipment working status and monitoring data. With reference to the LWM2M information model and resource definition, the equipment metadata, configuration parameters, and alarm events are modeled and described to realize the equipment remote configuration, firmware upgrade, and fault diagnosis management functions. Based on the condensation monitoring data set, a condensation risk assessment model is constructed. Support vector machine, random forest machine learning algorithm, LSTM, convolutional neural network deep learning algorithm are used to train and optimize historical monitoring data, and a nonlinear relationship between condensation risk level and monitoring indicators such as temperature and humidity, dew point temperature, and water vapor pressure is established. The model takes the real-time monitoring data of each monitoring point as input, outputs the condensation risk level of the point, and improves the accuracy and real-time performance of risk assessment through regular retraining and fine-tuning. According to the output results of the condensation risk assessment model, a substation condensation risk distribution map is dynamically generated to intuitively display the risk level and distribution of each monitoring point. Different colors of red, yellow and green are used to encode the risk level, and contour lines and heat maps are used to visualize the risk distribution. The substation geographic information layer and equipment layout layer are superimposed to focus on marking and analyzing high-risk areas, predict the time, location, and severity of condensation, and provide decision support for the subsequent optimization of protection strategies. Combined with the risk distribution map, environmental conditions, equipment performance factors, and the requirements for safe and stable operation of the power system, multi-objective optimization and reinforcement learning methods are used to dynamically optimize the existing condensation protection strategy of the substation. For the identified high-risk objects, extract their multi-dimensional attributes of environmental parameters, equipment parameters, and protection costs, and build protection strategy optimization models such as multi-objective linear programming and nonlinear programming. Set the objective function and weight coefficient of minimizing condensation risk and protection cost and maximizing equipment reliability, set environmental parameters, equipment parameters, and protection cost budget constraints, use NSGA-II and MOEA / D multi-objective optimization algorithms to solve, obtain the Pareto optimal solution set, select the optimal protection strategy parameters according to decision preferences, and conduct sensitivity analysis and robustness evaluation.The optimized protection strategy is sent to the dehumidification device, anti-ice flash coating spraying equipment, insulator cleaning equipment execution equipment and system. Through the edge computing gateway intelligent hardware, the strategy is converted into specific control instructions and action sequences, the equipment working parameters and operating status are adjusted in real time, and the strategy is closed-loop corrected and adaptively corrected according to the equipment execution feedback and monitoring data. Construct a condensation monitoring and protection intelligent operation and maintenance platform, use big data analysis, digital twin, virtual reality emerging technologies, and conduct full-process, full-scene, full-time simulation, monitoring, early warning, diagnosis and optimization of condensation monitoring, risk assessment, protection optimization, and equipment scheduling. Use big data analysis technology to conduct correlation mining and in-depth analysis of massive monitoring data, environmental data, and equipment data to extract condensation rules, risk patterns, and implicit knowledge of protection experience. Use digital twin technology to build a virtual simulation model of all elements, processes, and life cycles of substations to achieve realistic reproduction of monitoring scenarios, pre-rehearsal and deduction of risk events, and virtual and real verification of protection strategies. Use virtual reality technology to generate an immersive and interactive three-dimensional visualization interface, allowing managers to perceive condensation risks in person and command protection operations in real time. When realizing intelligent perception and monitoring of substations, intelligent sensors and acquisition devices that comply with the IEC61850 standard can be selected, such as intelligent temperature and humidity sensors, gas monitors, partial discharge monitors, etc., and sensor data can be uploaded to the substation monitoring system through standard protocols such as Modbus-TCP and IEC60870-5-104. For traditional analog sensors, intelligent terminals based on ARMCortex-M series chips and RS485 buses can be used to realize the digitization and networking transformation of analog signals. For equipment from different manufacturers, the IEC61970 / 61968-CIM common information model is used to model and describe the attributes, telemetry and telesignaling of the equipment, etc., to form a unified information exchange format and eliminate information islands. The data of each monitoring terminal is aggregated to the edge computing server of the substation control layer, and the MQTT message queue and Kafka data pipeline are used to realize cross-protocol forwarding, diversion storage, multi-level caching and other functions of data. We use the time series database InfluxDB and the containerized deployment tool Docker to build highly available and scalable data processing and storage nodes, and use the Kubernetes container orchestration platform to achieve elastic scheduling and fault self-healing of services. We update the collected multi-source heterogeneous data every 15 minutes, and ensure that the integrity and accuracy of the data reaches more than 99.9% through data fidelity compression, data quality exploration, and consistency verification.When conducting condensation risk assessment, the random forest algorithm is used to select features for environmental factors and equipment factors that affect condensation. Through grid search and cross-validation methods, the parameters of the random forest are optimized, including the number of decision trees, the maximum depth of each tree, and the maximum number of features considered when splitting nodes, etc., to obtain the feature importance ranking, and the top 10 features with a cumulative contribution rate of 95% are selected as model input. The input-output mapping model is constructed using a long short-term memory network, and environmental factors such as temperature, humidity, wind direction, wind speed, and air pressure and equipment operating status in the past 24 hours are used as inputs, and the condensation risk level in the next hour, such as low, medium, and high, is used as output to form a many-to-one time series prediction problem. A network structure consisting of 2 LSTM hidden layers and 1 Softmax output layer is built, using the Adam optimizer and cross entropy loss function, setting the learning rate to 0.001, the batch size to 64, the number of iterations to 500, and performing a model evaluation every 50 iterations. The accuracy rate on the training set reached 98%, the accuracy rate on the test set reached 95%, and the AUC value reached 0.99, indicating that the model can fit and predict condensation risk well. The model is retrained regularly every day to continuously improve the timeliness and robustness of the prediction. When optimizing the anti-condensation strategy, the optimization goals are to improve insulation reliability and reduce protection costs, the environmental factor threshold, the working parameters of the protection device, the operation plan frequency, etc. are used as optimization variables, and the condensation characteristic curve provided by the equipment manufacturer and the protection experience of the operation and maintenance personnel are used as constraints to establish a multi-objective mixed integer programming model. The model is solved using an improved non-dominated sorting genetic algorithm, with the population size set to 100, the crossover probability to 0.8, the mutation probability to 0.1, the number of iterations to 500, and the number of Pareto optimal solutions to 20. The algorithm is implemented in the Gurobi optimization library, and the algorithm is accelerated using a high-performance computing cluster, shortening the solution time from hours to minutes. Taking the protection cost and reliability level as evaluation indicators, the Pareto optimal solution set is quantitatively evaluated and sorted, and the solution with the best comprehensive performance is automatically selected as the final protection strategy. Simulation experiments show that the optimized strategy can reduce the equipment failure rate from 2% to 0.5%, and reduce the annual protection cost from 500,000 yuan to 300,000 yuan, achieving a double improvement in reliability and economy. When building the intelligent operation and maintenance platform, a microservice architecture with front-end and back-end separation is adopted, and mainstream frameworks such as SpringBoot and Vue.js are used to develop core functional modules such as monitoring data management, condensation risk assessment, protection strategy optimization, intelligent alarm decision-making, equipment ledger management, and three-dimensional visualization, and the REST interface is used to realize data interaction between modules. Using streaming computing frameworks such as Storm and SparkStreaming, the monitoring data is cleaned, filtered, aggregated and analyzed in real time, and Redis cache and RabbitMQ message queue are used to achieve high-speed data access and cross-platform sharing.Use visualization engines such as ECharts and BabylonJS to build cool 2D and 3D interfaces, and use immersive technologies such as VR / AR to create an immersive virtual operation and maintenance experience. Use Neo4j graph database and Cypher query language to build a semantic network of knowledge in the fields of equipment, environment, strategy, etc., and realize graph-based knowledge reasoning and intelligent question and answer. Use Drools rule engine and BPMN workflow engine to solidify business processes into computer-executable rules and processes to achieve automation and standardization of decision-making. Use digital twin simulation platform to build a multi-physics field simulation model of substation, carry out refined simulation of typical scenarios such as equipment aging, material corrosion, condensation accumulation, etc., and optimize sensor deployment plan and inspection and maintenance strategy. Use OneMapGIS platform to integrate condensation monitoring data with geospatial data to generate substation condensation risk map and protection resource map, and assist managers in visual analysis and scientific decision-making.

Claims

1. A condensation monitoring method for a digital substation, comprising the following steps: S1. Obtain data information of the target digital substation; S2. According to the data information obtained in step S1, based on the numerical simulation scheme, simulate the airflow movement characteristics in the target digital substation, determine the high-risk areas for condensation, and arrange corresponding monitoring equipment to obtain monitoring data; S3. According to the difference in the response characteristics of electromagnetic waves in different frequency bands to condensation, a monitoring frequency band is selected, and data of the monitoring frequency band is obtained and analyzed, so as to construct a feature set related to the degree of condensation; S4. Obtain the monitoring data of the target digital substation in real time, and select the electromagnetic wave monitoring frequency band based on the current environment; S5. According to the electromagnetic wave monitoring frequency band obtained in step S4, feature extraction and analysis are performed to construct a condensation monitoring model based on electromagnetic wave transmission characteristics; S6. Optimizing the condensation monitoring model constructed in step S5 based on the fused environmental parameters and electromagnetic wave parameters; S7. Based on the optimized condensation monitoring model obtained in step S6, a condensation risk prediction result of the target digital substation is generated; S8. According to the prediction result obtained in step S7, the optimal anti-condensation measures are given to the target digital substation, and the condensation monitoring of the target digital substation is completed.

2. The condensation monitoring method for a digital substation according to claim 1 is characterized in that The step S2 specifically includes the following steps: A three-dimensional model of the target digital substation is established, and the airflow movement in the target digital substation is numerically simulated using a computational fluid dynamics solution to obtain the characteristic parameters of the airflow movement in different areas of the target digital substation; The obtained air flow motion characteristic parameters are correlated with temperature and humidity to determine the high-risk areas for condensation in the target digital substation; In the identified high-risk areas, corresponding monitoring equipment is deployed to obtain monitoring data.

3. The condensation monitoring method for a digital substation according to claim 2 is characterized in that The step S3 specifically includes the following steps: Based on the electromagnetic wave transmission theory and the analysis of the physical mechanism of condensation, several frequency bands that are sensitive to condensation are selected as candidate frequency bands; Build a condensation simulation experimental platform, conduct experiments under different condensation conditions, and obtain electromagnetic wave transmission characteristic test data of candidate frequency bands; Fitting and regression analysis were performed on the obtained electromagnetic wave transmission characteristic test data to obtain the correlation scale model between the condensation degree and the electromagnetic wave frequency band and transmission characteristic parameters; According to the correlation scale model, the monitoring frequency band is determined to obtain a monitoring frequency band set; According to the obtained monitoring frequency band, electromagnetic wave transmission data is collected to obtain an electromagnetic wave transmission data set reflecting the condensation state; Perform video domain analysis on the electromagnetic wave transmission data reflecting the condensation state, and extract characteristic parameters that can reflect the degree of condensation; The feature selection algorithm is used to screen and optimize the obtained feature parameters, obtain the optimal feature subset related to the condensation degree, and construct the condensation degree prediction feature set; The obtained condensation degree prediction feature set is used to train the support vector regression algorithm to obtain a condensation degree prediction model; The obtained condensation degree prediction model is used to predict the condensation degree of the target digital substation.

4. The condensation monitoring method for a digital substation according to claim 3 is characterized in that The step S4 specifically comprises the following steps: Real-time acquisition of environmental monitoring data of the target digital substation; the environmental monitoring data includes temperature data, humidity data, pressure data and light intensity data of the UB Aokuo monitoring point; Clustering algorithm is used to perform clustering analysis on the acquired environmental monitoring data, and the target digital substation is divided into several environmental areas according to the clustering results; Install condensation sensors in various environmental areas to monitor condensation data on the equipment surface in real time; Select a monitoring frequency band and use the selected monitoring frequency band to transmit condensation data data.

5. The condensation monitoring method for a digital substation according to claim 4 is characterized in that The step S5 specifically includes the following steps: Based on the maximum relevance minimum redundancy algorithm and the Relief-F algorithm, feature selection is performed on the electromagnetic wave attenuation feature data, and the attenuation feature with the greatest correlation with the condensation degree is screened out as the candidate feature set. Feature extraction and fusion are performed on the candidate feature set. The principal component analysis method is used to extract the main component features of the attenuation feature. The independent component analysis method is used to extract the independent component features of the attenuation feature. The autoencoder is used to extract the deep nonlinear features of the attenuation feature to obtain the attenuation feature parameters. Acquire condensation image data and video data on the device surface; Extract image feature parameters of the acquired image data and video data, score the condensation degree according to the set requirements, and obtain a condensation observation sample data set; The condensation observation sample data set is used to train the gradient boosting regression tree model to obtain the condensation monitoring model. Obtain the weather forecast data, internal environment monitoring data and electromagnetic wave attenuation data of the target digital substation, and use the obtained data to train the long short-term memory neural network to obtain the condensation trend prediction model; The condensation trend prediction model is used to predict the condensation trend of the target substation.

6. The condensation monitoring method for a digital substation according to claim 5 is characterized in that The step S6 specifically includes the following steps: The acquired environmental parameters and electromagnetic wave parameters are preprocessed through data alignment, time series interpolation and outlier removal algorithms to obtain fused feature vectors; The condensation monitoring model constructed in step S5 is optimized by using the fused feature vector.

7. The condensation monitoring method for a digital substation according to claim 6 is characterized in that The step S7 specifically includes the following steps: The optimized condensation monitoring model is used to predict the condensation risk of the target digital substation; According to the predicted condensation risk probability, clustering algorithm is used to perform clustering and grouping, and different risk levels are marked; The monitoring grid of the target digital substation is set as an undirected weighted graph, each monitoring point corresponds to a node of the graph, there is a weighted edge between adjacent monitoring points, and the weight of the edge is set; The Node2Vec algorithm is used to optimize the node collinearity probability to obtain the low-dimensional embedding vector of the node; the IDW algorithm is used to interpolate and predict the risk level of the unknown node; based on the interpolation prediction results, the substation geographic information map is used as the base map and the monitoring points are used as control points to generate the risk distribution map; Areas with different risk levels are marked.

8. The condensation monitoring method for a digital substation according to claim 7 is characterized in that The step S8 specifically includes the following steps: According to the prediction result obtained in step S7, the area where the condensation risk is greater than the set threshold is obtained, and the specific location and coverage are determined in combination with the layout data of the target substation to obtain a location list of high-risk areas for condensation; Build a knowledge base of anti-condensation measures; In the anti-condensation measures knowledge base, the environmental characteristics of the high-risk area for condensation are compared with the environmental characteristics data in the anti-condensation measures knowledge base for similarity, and the anti-condensation measures with similarity higher than the set threshold are selected as candidate measures; Among the candidate measures, the cost data of each candidate measure is calculated according to the installation location and installation method of the candidate measure, and the changes in regional environmental parameters after the implementation of each candidate measure are calculated through the three-dimensional model of the target digital substation, and the anti-condensation effect of the candidate measures is evaluated. According to the cost data and the anti-condensation effect, the final anti-condensation measures are determined to complete the condensation monitoring of the target digital substation.

9. The condensation monitoring method for a digital substation according to any one of claims 1 to 8, characterized in that: The following steps are also included: S9. Establish a condensation monitoring and protection platform to achieve condensation monitoring, data management and optimization of protection measures for the target digital substation.

10. The condensation monitoring method for a digital substation according to claim 9, characterized in that The step S9 specifically includes the following steps: Perform data preprocessing on the acquired data to generate a condensation monitoring data set; Based on support vector machine and convolutional neural network, an initial model for condensation risk assessment is constructed, and the generated condensation monitoring data set is used for training to obtain a condensation risk assessment model. The obtained condensation risk assessment model is used to obtain the condensation risk assessment result, and combined with the three-dimensional model of the target digital substation, a condensation risk distribution map is generated; According to the generated condensation risk distribution map, the anti-condensation measures are dynamically optimized based on a multi-objective planning scheme.

Images